Solution to So You Think You Can Count?
by Charles Tam
This puzzle presents as a simple chat room with a stated goal: All your team has to do is count to 100, one by one, without skipping or repeating any numbers. This is a traditionally difficult coordination exercise in live chats worldwide. To better simulate the posting conditions to which solvers may be accustomed, we have provided our own trolls to assist with the task (since we know that none of our solving teams contain Internet trolls at all). The moment any participant successfully counts 100, the page reveals the correct answer, YOSHIMITSU.
The behavior of the six automated trolls is described below, in the order of their expected appearance. Their names have no special significance except to hint at their behavior.
Talk Like an Animal
As an initial layer of difficulty, all participants in chat are assigned a random animal name to better disguise the presence of our automated trolls. We hope you all enjoyed telling off whichever of your human team members you thought was the Helpful Firefly.
Helpful Firefly announces its presence after some number N with a suitably human-like message. It then counts N+1 and, three seconds later, N+3. This tests the solving team's ability to designate exactly one person to submit N+2 in the small window of reaction time given.
In addition, teams that are too efficient in their counting may find that all their active members are on the five-second message cooldown when Helpful Firefly appears.
Singing Creodont is not a standalone entity, but instead replaces all chat text with some number of A's. This is a purely cosmetic effect — numeric inputs are still processed as normal, and replaced by that number of A's. While solvers can technically keep counting and ignore the Creodont, its yelling may interfere with their ability to recognize when one of the other trolls is in play.
Singing Creodont's effect can be disabled by singing along, providing the correct number of A's (e.g. 13, in the case of Ember Husky above). Providing an incorrect number of A's registers as a counting failure.
Quizb Owl asks relatively simple trivia questions. The problem is that their answers are all numbers, so solvers have to write them out in words to avoid a miscount. Questions attempt to induce one of several failure modes — timeout failure because the spelled-out version of the number is long; communication failure because the answer is chosen to be in the vicinity of a number that is the correct count; and brain failure because the question nerd snipes hard enough for the answerer to forget about the whole use-words thing.
Quizb Owl insists upon an answer within forty seconds, posting increasingly agitated reminders every ten seconds. At the forty second mark, Quizb Owl posts the correct answer, as a number, which may trigger a miscount.
Adder Viper tries to backsolve this puzzle through a long and complicated simulation process that is equivalent to summing the two components of its version number. Solvers need to race with the Viper's startup process to make sure that their expected count is at the right place when the Viper spits out its projected answer to the problem.
The amount of time allotted, and the Viper's eventual answer, are chosen so that a team of four will find the race to be fast but not impossibly so. Three people can manage it as long as they can dispatch other interfering bots easily. Two people have to play perfectly and might still be shut out by a bad roll (keeping in mind that each person can only post once every five seconds).
Twitchy Chat, much like its namesake, just wants solvers to know that it is good at solving puzzles too. It is sometimes correct and sometimes wrong, but never actually triggers a count or miscount.
Nice Ouroboros just wants to post the meme number.